Dissemination of lipid A deacylases (PagL) among Gram-negative bacteria IDENTIFICATION OF ACTIVE-SITE HISTIDINE

Lipopolysaccharide (LPS) is one of the main constituents of the Gram-negative bacterial outer membrane. It usually consists of a highly variable O-antigen, a less variable core oligosaccharide, and a highly conserved lipid moiety, designated lipid A. Several bacteria are capable of modifying their lipid A architecture in response to external stimuli. The outer membrane-localised lipid A 3- O -deacylase, encoded by the pagL gene of Salmonella enterica serovar Typhimurium, removes the fatty acyl chain from the 3 position of lipid A. Although a similar activity was reported in some other Gram-negative bacteria, the corresponding genes could not be identi ﬁ ed. Here, we describe the presence of pagL homologs in a variety of Gram-negative bacteria. Although the overall sequence similarity is rather low, a conserved domain could be distinguished in the C-terminal region. The activity of the Pseudomonas aeruginosa and Bordetella bronchiseptica pagL homologs was con ﬁ rmed upon expression in Escherichia coli , which resulted in the removal of an R -3-hydroxymyristoyl group from lipid A. Upon deacylation by PagL, E. coli lipid A underwent another modi ﬁ cation, which was the result of the activity of the endogenous palmitoyl transferase PagP. Furthermore, we identi ﬁ ed a conserved histidine-serine couple as -residues, suggesting a catalytic mechanism similar to serine hydrolases. The biological function of PagL remains unclear. However, because PagL homologs were found in both pathogenic and non-pathogenic species, PagL-mediated deacylation of lipid A probably does not have a dedicated role in pathogenicity. contain enzymes with an activity similar to that of PagL. We report now the identi ﬁ cation of PagL homologs in a variety of Gram-negative bacteria. The limited sequence similarity among the various proteins was used to identify active-site residues.


Introduction
Lipopolysaccharide (LPS), a major component of the Gram-negative bacterial outer membrane, is known to be important for the functioning of this membrane as a permeability barrier and for the resistance against complement-mediated cell lysis (for review, see Raetz and Whitfi eld, 2002). It consists of three covalently linked domains: lipid A, the core, and the O-antigen. Lipid A forms the hydrophobic membrane anchor and is responsible for the endotoxic activity of LPS. In Escherichia coli, it consists of a 1, 4'-bisphosphorylated β-1,6-linked glucosamine disaccharide, which is replaced by R-3-hydroxymyristic acid residues at positions 2, 3, 2', and 3' via ester or amide linkage. Secondary lauroyl and myristoyl groups replace the hydroxyl group of R-3hydroxymyristoyl at the 2' and 3' positions, respectively (Fig. 1A). Previous studies have shown that the phosphate groups, the glucosamine disaccharide, and the correct number and length of the acyl chains are important for the biological activity of lipid A (Raetz and Whitfi eld, 2002;Loppnow et al., 1989;Steeghs et al., 2002). The basic structure of lipid A is reasonably well conserved among Gram-negative bacteria, although slight variations in the pattern of the substitutions of the two phosphates and the acyl chain number and length are observed (Nikaido and Vaara, 1987;Caroff et al., 2002).
Additional modifi cations of lipid A (Fig. 1B) are regulated in Salmonella enterica serovar Typhimurium (S. Typhimurium) by the two-component regulatory system PhoP/PhoQ (Guo et al., 1997;Guo et al., 1998). In response to low Mg 2+ levels, the sensor kinase PhoQ phosphorylates and thereby activates the transcriptional activator PhoP, which leads to the activation or repression of 40 different genes (Guo et al., 1997;Gunn et al., 1998a). A second regulatory system involved in lipid A modifi cation is the PmrA/ PmrB two-component system, which itself is PhoP/PhoQ-regulated (Gunn et al., 1998b;Gunn et al., 2000). Mutants with alterations in the PhoP/PhoQ system exhibit reduced virulence and an increased susceptibility to anti-microbial peptides (Miller et al., 1989;Gunn and Miller, 1996). Homologs of the PhoP/PhoQ and PmrA/PmrB systems have been identifi ed in other Gram-negative bacteria, including E. coli, Yersinia pestis, and Pseudomonas aeruginosa (Ernst et al., 1999a;Ernst et al., 1999b).
Until now, several lipid A-modifying enzymes have been identifi ed (Fig. 1B). Substitution of the 1-and 4'-phosphate groups with one or two 4-amino-4-deoxy-Larabinose (L-Ara4N) moieties in S. Typhimurium was found to be dependent on the enzyme ArnT (Trent et al., 2001b). Recently, the PmrC protein was identifi ed to mediate the addition of phosphoethanolamine to lipid A in S. enterica (Lee et al., 2004). Another enzyme, designated LpxO, catalyses the O 2 -dependent hydroxylation of lipid A , and a lipid A 1-phosphatase was identifi ed in Rhizobium leguminosarum coli lipid A consists of a bisphosphorylated glucosamine disaccharide substituted with four R-3-hydroxymyristoyl moieties, of which the 2'-and 3'-fatty-acyl chains are esterifi ed with laurate and myristate, respectively. (B) regulated modifi cations of Salmonella lipid A. Substitution of the phosphate moieties with L-Ara4N or phosphoethanolamine is mediated by ArnT and PmrC, respectively, the formation of a 2-hydroxymyristate-modifi ed lipid A by LpxO, the addition of a secondary palmitoyl chain at the 2 position by PagP, and the removal of the 3-hydroxymyristoyl moiety at the 3 position by PagL are shown. (Karbarz et al, 2003). All these enzymes are thought to reside within the inner membrane or periplasmic space (Trent et al., 2001b;Lee et al., 2004;Gibbons et al., 2000;Karbarz et al., 2003). Recently, a new class of outer membrane-localised lipid A-modifying enzymes was discovered. One of them is the palmitoyl transferase PagP (Bishop et al., 2000). Palmitoylation of lipid A leads to an increased resistance to cationic antimicrobial peptides (Guo et al., 1998). Furthermore, palmitoylated lipid A antagonises LPS-induced activation of human cells (Tanamoto and Azumi, 2000). Homologs of PagP are found, among others, in S. Typhimurium, Bordetella pertussis, Bordetella bronchiseptica, Bordetella parapertussis, Legionella pneumophila, E. coli, and Y. pestis (Bishop et al., 2000;Robey et al., 2001). Another outer membrane-localised lipid A-modifying enzyme is the 3-O-deacylase PagL (Trent et al., 2001a). This enzyme was discovered in S.
Typhimurium and shown to hydrolyse the ester bond at the 3 position of lipid A, thereby releasing the primary 3-hydroxymyristoyl moiety (Trent et al., 2001a). At that time, no obvious homologs of this protein could be found in the nonredundant or unfi nished microbial databases, except in the closely related strains S. enterica serovars Typhi and Paratyphi (Trent et al., 2001a). Nevertheless, some other Gram-negative bacteria, including P. aeruginosa (Ernst et al., 1999b), R. leguminosarum (Bhat et al., 1994), Helicobacter pylori (Moran et al., 1997), and Porhyromonas gingivalis (Kumada et al., 1995) contain 3-O-deacylated lipid A species, suggesting that these organisms contain enzymes with an activity similar to that of PagL. We report now the identifi cation of PagL homologs in a variety of Gram-negative bacteria. The limited sequence similarity among the various proteins was used to identify active-site residues.

Bacterial strains and growth conditions
All bacterial strains used in this study are described in Table 1. Unless otherwise notifi ed, the E. coli and P. aeruginosa strains were grown at 37 o C and 30 o C, respectively, in a modifi ed Luria-Bertani broth, designated LB (Tommassen et al., 1983), supplemented with 0.2% glucose, or in minimal medium (SV) (Winkler and de Haan, 1949) supplemented with 0.5% glucose, while shaking at 200 rpm. To induce expression of the pagL genes cloned behind the T7 promoter, the bacteria were grown in LB supplemented with glucose until an absorbance at 600 nm (A 600 ) of 0.4-0.6 was reached. Expression of the pagL genes was then induced by adding 1 mM isopropyl-1-thio-β-D-galactopyranoside (IPTG), and incubation at 37 o C was continued. When appropriate, bacteria were grown in the presence of 100 μg/ml ampicillin, 50 μg/ml kanamycin, 10 μg/ml tetracycline, or   (Pa) pET-11a derivative harboring P. aeruginosa pagL This study pPagL (Bb) pET-11a derivative harboring B. bronchiseptica pagL This study pPagL (St) pET-11a derivative harboring S. Typhimurium pagL This study pPagL (Pa) (-) pET-11a derivative encoding P. aeruginosa PagL without signal sequence This study pPagL (Pa)  The pagL genes from S. Typhimurium SR11 (pagL (St) ), B. bronchiseptica B505 (pagL (Bb) ), and P. aeruginosa PAO25 (pagL (Pa) ) were cloned into pET-11a (Novagen) behind the T7 promoter. The genes were amplifi ed by PCR using chromosomal DNA as template. Template DNA was prepared by resuspending ~10 9 bacteria in 50 μl of distilled water, after which the suspension was heated for 15 min at 95 o C. The suspension was then centrifuged for 1 min at 16,100 x g, after which the supernatant was used as template DNA. The sequences of the forward primers, which contained an NdeI site (underlined), including an ATG start codon, were 5'-AACATATGAAGAGAATATTTATATATC-3' (pagL (St) ), 5'-AACATATGAAGAAACTACTTCCGCTGG-3' (pagL (Pa) ), and 5'-AACATATGCAATTTCTCAAGAAAAACA-3' (pagL (Bb) ). The sequences of the reverse primers, which contained a BamHI site (underlined) and included a stop codon, were 5'-AAGGATCCTCAGAAATTATAACTAATT-3' (pagL (St) ), 5'-AAGGATCCCTAGATCGGGATCTTGTAG-3' (pagL (Pa) ), and 5'-AAGGATCCTCAGAACTGGTACGTATA-G-3' (pagL (Bb) (Sambrook et al., 1989). Plasmid DNA from transformants was checked for presence of the correct PagL-encoding insert by digestion with NdeI and BamHI. Plasmids that gave a correct digestion profi le were designated pPagL (Pa) , pPagL (Bb) , and pPagL (St) ( Table 1). The correct coding sequences of the cloned pagL genes were confi rmed by nucleotide sequencing in both directions. Mutations were introduced in pagL by using the QuikChange site-directed mutagenesis kit (Stratagene) GCCGGCGTTGGAATAGTTGATCGCCCGAACGCC S151A_FW CGGGCGATCCACTATGCGAACGCCGGCCTGAAA S151A_REV TTTCAGGCCGGCGTTCGCATAGTGGATCGCCCG S151C_FW CGGGCGATCCACTATTGCAACGCCGGCCTGAAA S151C_REV TTTCAGGCCGGCGTTGCAATAGTGGATCGCCCG a The primer name gives the amino acid substitution, e.g. H81A_FW indicates that the oligonucleotide shown was used as the forward primer in a site-directed mutagenesis procedure to substitute the histidine at position 81 of the precursor PagL (Pa) by an alanine. b Introduced mutations are underlined. and the primers listed in Table 2. Plasmid pPagL (Pa) was used as the template in which the mutations were created. The presence of the correct mutations was confi rmed by nucleotide sequencing in both directions.

Isolation of cell envelopes
Cells were harvested by centrifugation for 10 min at 1,500 x g and washed once in 50 ml of cold 0.9% sodium chloride solution. The cell pellets were frozen for at least 15 min at -80 o C and then suspended in 20 ml of 3 mM EDTA, 10 mM Tris-HCl (pH 8.0) containing Complete protease inhibitor mixture (Roche Applied Science). The cells were disrupted by sonication, after which unbroken cells were removed by centrifugation for 10 min at 1,500 x g. The cell envelopes were pelleted from the supernatant by centrifugation for 1.5 h at 150,000 x g and resuspended in 2 mM Tris-HCl (pH 7.4). The cell envelopes were stored at -80 o C in aliquots.

SDS-PAGE and immunoblotting
Proteins were analysed by SDS-PAGE (Laemmli, 1970) with 0.2% SDS in the running gel using the Bio-Rad Mini-PROTEAN ® 3 apparatus. Samples were applied to a 13% polyacrylamide gel with a 4% stacking gel and subjected to electrophoresis at 150 V.

Proteins were stained with Coomassie Brilliant Blue. Prestained or unstained Precision
Plus Protein TM Standard from Bio-Rad was used to determine the relative molecular mass. For Western blotting, proteins were transferred from SDS-polyacrylamide gels onto nitrocellulose membranes. The membranes were blocked overnight in phosphatebuffered saline (pH 7.6), 0.5% nonfat dried milk, 0.1% Tween 20 and incubated with guinea pig antibodies directed against PagL (Pa) in blocking buffer followed by an incubation with horseradish peroxidase-conjugated rabbit anti-guinea pig IgG antibodies (Sigma) in blocking buffer. Blots were developed using SuperSignal ® WestPico Chemiluminescent Substrate (Pierce).

Polyclonal antibodies
For antibody production, the pagL gene from P. aeruginosa PAO25 without the signal sequence-encoding part was PCR amplifi ed by using the forward primer (5'-AACATATGGCGGACGTCTCGGCCGCCG-3'), which contained an NdeI site (underlined), including an ATG start codon, and the reverse primer (5'-AAGGATCCCTAGATCGGGATCTTGTAG-3'), which contained an BamHI site (underlined) and included a stop codon. The PCR product was cloned into pET-11a, and the resulting plasmid, pPagL (Pa) (-) , was used to transform E. coli BL21 Star TM (DE3) to allow for expression of the truncated pagL gene. The PagL (Pa) protein, accumulating in inclusion bodies, was isolated (Dekker et al., 1995), purifi ed from a preparative SDSpolyacrylamide gel, and used for immunisation of guinea pigs at Eurogentec.

Microsequencing
Proteins were transferred from SDS-polyacrylamide gels to an Immobilon-P polyvinylidene difl uoride membrane (Millipore Corp.) in 192 mM glycine, 25 mM Tris (pH 8.3), 10% methanol (v/v) at 100 V for 1 h using the Bio-Rad Mini-PROTEAN ® 2 blotting apparatus. After transfer, the membrane was washed three times for 15 min with distilled water. Transferred proteins were stained with Coomassie Brilliant Blue.
The membrane was dried in the air, and the putative PagL bands were excised and subjected to microsequencing at the Sequencing Center Facility, Utrecht University, The Netherlands.

LPS analysis by Tricine-SDS-PAGE
Approximately 10 9 bacteria were suspended in 50 μl of sample buffer (Laemmli, 1970), and 0.5 mg/ml proteinase K (end concentration) was added. The samples were incubated for 60 min at 55 o C followed by 10 min at 95 o C to inactivate proteinase K.
The samples were then diluted 10-fold by adding sample buffer, after which 2 μl of each sample was applied to a Tricine-SDS-polyacrylamide gel (Lesse et al., 1990).
The bromphenol blue was allowed to run into the separating gel at 35 V, after which the voltage was increased to 105 V. After the front reached the bottom of the gel, the samples were left running for another 45 min. The gels were fi xed overnight in water/ ethanol/acetic acid 11:8:1 (v/v/v) and subsequently stained with silver as described previously (Tsai and Frasch, 1982).
Gas chromatography-mass spectrometry (GC/MS) and electrospray ionisationmass spectrometry (ESI/MS) LPS was isolated using the hot phenol/water extraction method (Westphal and Jann, 1965). Structural analysis of purifi ed lipid A was performed by nanoelectrospray tandem MS on a Finnigan LCQ in the negative ion mode (Wilm and Mann, 1996).

Identifi cation of PagL homologs in various Gram-negative bacteria
The 187-amino acid sequence of the S. Typhimurium PagL precursor protein (GenBank accession no. AAL21147) was used as a lead to identify putative PagL homologs in other Gram-negative bacteria, by searching all completed and unfi nished genomes of Gram-negative bacteria present in the NCBI data base (www.ncbi.nlm. nih.gov/sutils/ genom_table.cgi). BLAST search (Altschul et al., 1990) revealed the presence of putative homologs in the Bordetella spp. B. pertussis, B. bronchiseptica, and B. parapertussis (Fig. 2). The PagL homologs of B. bronchiseptica and B. parapertussis are two mutually identical 178-amino acid polypeptides ( Fig. 2) with, as predicted by the SignalP server (Nielsen et al., 1999), a 25-amino acid N-terminal signal peptide. A gene for a PagL homolog was also found in the genome of the B. pertussis Tohama I strain (Parkhill et al., 1999), but this open reading frame ( (Fig. 2). Together, all PagL homologs exhibited a low overall mutual sequence identity but contained a clear homologous domain near the C terminus.

Cloning of pagL and heterologous expression in E. coli
To verify their putative lipid A deacylase activity, we cloned the pagL homologs of P. aeruginosa (pagL (Pa) ) and B. bronchiseptica (pagL (Bb) ). We included in these studies pagL from S. Typhimurium (pagL (St) ) as a reference. These pagL genes were amplifi ed from the chromosomes by PCR and eventually cloned into pET-11a under the control of the T7 promoter, resulting in plasmids, pPagL (Pa) , pPagL (Bb) , and pPagL ( respectively. Particularly in the case of expression of PagL (Bb) , an additional band with a higher molecular mass was visible on the gel (Fig. 3).The N-terminal sequence of this band, MQFLK, corresponded with that of the precursor of PagL (Bb) .

In vivo modifi cation of E. coli LPS by PagL
To study whether the cloned PagL homologs were active on E. coli LPS, IPTG was added to exponentially growing E. coli BL21 Star TM (DE3) cells containing the empty vector pET-11a or the pPagL plasmids, and after various incubation periods, samples equivalent to one A 600 unit were collected, and their LPS content was analysed by Tricine-SDS-PAGE. In accordance with the expected hydrolysis of the R-3-hydroxymyristate at the 3 position of lipid A, expression of any of the three pagL homologs converted the LPS into a form with a higher electrophoretic mobility (Fig. 4). The conversion was almost complete within 75 min after PagL (Pa) or PagL (Bb) was induced but took somewhat longer in the case of PagL (St) . To test whether the effi ciency of in vivo modifi cation in E. coli was dependent on growth temperature (30, 37, or 42 o C), the presence of magnesium chloride (10 mM), time of PagL induction during the growth phase (early log phase, mid log phase, or stationary phase), or nutrient availability (rich medium (LB), or minimal medium (SV)), PagL (Pa) was expressed in E. coli under various conditions, after which the LPS profi le was analysed by Tricine-SDS-PAGE. Under all conditions tested, no obvious changes in deacylation effi ciency could be observed (data not shown).

Structural analysis of PagL-modifi ed LPS
To determine its fatty acid content, LPS was isolated from bacteria that were grown in the presence of 10 mM MgCl 2 to suppress PhoP/PhoQ-regulated modifi cations of lipid A and analysed by GC/MS. The C14/3OH C14 ratio in the PagL-modifi ed LPS samples was increased compared with that in the wild-type LPS (Fig. 5), consistent with the expected removal of a 3OH C14 from lipid A. To confi rm these data, the lipid A moieties were isolated and analysed by ESI/MS in the positive ion mode, which revealed the presence of four major lipid A species in wild-type LPS (Fig. 6A). The peak at m/z 1797 represents the characteristic hexaacylated bis-phosphate species that is typically found in E. coli, whereas the peak at m/z 1928 corresponds to a hexaacylated bis-phosphate species replaced with an L-Ara4N moiety. The two remaining peaks at m/z 1716 and m/z 1847 most likely represent fragment ions of the two former species missing a phosphate group. Upon expression of PagL (St) (Fig. 6B), PagL (Pa) (Fig. 6C), or PagL (Bb) (Fig. 6D), the major lipid A species were present at m/z 1622 and m/z 1490, which correspond to the loss of one β-hydroxymyristate residue and one phosphate group from the major species at m/z 1928 and m/z 1797 present in the empty vector control, respectively. Also here, the loss of the phosphate group is probably an artifact of the ionisation procedure. Based upon the GC/MS and ESI/MS data, it can be concluded that the identifi ed PagL homologs of P. aeruginosa and B. bronchiseptica, like that of S. Typhimurium, are active lipid A deacylases. Furthermore, the data suggest that the deacylation is not dependent upon the absence or presence of an L-Ara4N moiety because both species were deacylated effi ciently.

Subsequent in vivo modifi cation of PagL-deacylated LPS
In the course of these experiments, it was observed that after prolonged PagL expression, PagL-modifi ed LPS was no longer detectable on Tricine-SDS-PAGE and that the LPS migrated again at the position of wild-type LPS, as illustrated for the strain Exponentially growing E. coli BL21 Star TM (DE3) cells containing pET-11a or the pPagL constructs were induced with IPTG for the indicated time, after which 1 A 600 unit culture samples were collected and analysed by Tricine-SDS-PAGE. expressing PagL (Bb) (Fig. 7A). The PagL protein was still abundantly present at this time point, as revealed on SDS-PAGE (data not shown). Furthermore, analysis by GC/MS revealed that the C14/3OH C14 ratio was not decreased again for the LPS isolated after a 5-h induction of PagL (Bb) (Fig. 7B). Thus, the secondary modifi cation observed on the Tricine-SDS-polyacrylamide gel (Fig. 7A) was not the consequence of restoration of the PagL modifi cation, but the result of (an) additional modifi cation(s) that restored the electrophoretic mobility to that of wild-type LPS. Therefore, other fatty acid ratios were compared. A striking increase in the C16/C14 ratio was found in the LPS of cells induced 5 h for PagL production (Fig. 7C), suggesting that the PagL-deacylated LPS was subsequently palmitoylated.
A protein that adds palmitate to lipid A is the outer membrane protein PagP (Bishop et al., 2000) (Fig. 1). Therefore, we hypothesised that the secondary modifi cation of PagL-modifi ed LPS might have been the result of endogenous PagP activity. To investigate this possibility, we transformed wild-type E. coli BL21 Star TM (DE3) and its pagP mutant derivative JG101 with the pPagL (Pa) plasmid. The secondary modifi cation of PagL-modifi ed LPS was again observed in the case of the wild-type strain, but not in that of the mutant strain (Fig. 7D). This result strongly suggests that the secondary modifi cation of PagL-modifi ed LPS (Fig. 7A) was indeed the consequence of endogenous PagP activity.

Identifi cation of PagL active-site residues
The mutual sequence identity between the identifi ed PagL homologs is very low (Fig. 2). Among the few totally conserved residues are a histidine and a serine, which, we hypothesise, might be part of a "classical" Asp/Glu-His-Ser catalytic triad of serine hydrolases. These putative active-site residues are located at the lipid-exposed side near the top of a β-strand in a topology model we propose (Fig. 8). Interestingly, in the outer membrane phospholipase A (OMPLA), the active-site His and Ser are located in a similar position (Snijder et al., 1999). To test whether these residues, located at positions 149 and 151 of the PagL (Pa) precursor protein, respectively, are indeed important for catalytic activity, they were replaced by alanine or asparagine, and by alanine or cysteine, respectively. As a control, the same substitutions were made for a non-conserved was no longer observed when the conserved His-149 and Ser-151 were replaced (Fig.   9B), even though the expression of these mutant proteins was not affected (Fig. 9A).
These results strongly support the hypothesis that the conserved histidine at position 149 and serine at position 151 of the precursor PagL (Pa) protein are active-site residues and that PagL mechanistically functions as a serine hydrolase.

PagL expression in P. aeruginosa and phenotypic characterisation of a pagL mutant
Because pagL homologs were identifi ed in many Gram-negative bacteria, including non-pathogenic soil bacteria (Fig. 2), a primary role for the enzyme in pathogenesis appears unlikely. To gain insight in the possible function of PagL in P. Residues in the postulated β-strands are shown in diamonds, which are shaded for residues that are exposed to the lipid bilayers. His-149 and Ser-151 (marked in red; position in the PagL (Pa) precursor) are absolutely conserved (Fig. 2) and are suggested to be part of a classical catalytic triad of a serine hydrolase. Potential candidates for the acidic residue of the catalytic triad are indicated in yellow. Numbers refer to the positions of the residues in the precursor sequence.
aeruginosa, we tested whether endogenous pagL expression levels in P. aeruginosa can be infl uenced by the growth conditions. Therefore, wild-type P. aeruginosa PAO1 was grown in LB supplemented or not with 0-5% ethanol, 10 mM magnesium chloride, or 5 μg/ml EDDHA (to create iron-limiting conditions), or in minimal medium, or on solidifi ed media (LB agar, and SV agar). Endogenous PagL levels were analysed by immunoblotting. The result showed that under all conditions tested, PagL was expressed, and the expression levels were similar (data not shown). Furthermore, we compared the growth characteristics of a P. aeruginosa pagL transposon-insertion mutant (32751) and its parental strain (PAO1). Both strains were tested for their ability to grow in LB medium at different temperatures (25, 30, or 37 o C) or in LB medium supplemented with 0-4 M Fig. 9 Identifi cation of PagL (Pa) active-site residues by amino acid substitution. Exponentially growing E. coli BL21 Star TM (DE3) cells containing the empty pET-11a vector, the pPagL (Pa) plasmid, or the mutant pPagL (Pa) plasmids were induced with IPTG for 75 min, after which 1 A 600 unit culture samples were collected and analysed by SDS-PAGE followed by immunoblotting with primary antibodies against PagL (Pa) (A) and by Tricine-SDS-PAGE to visualise LPS (B). sodium chloride, 0-1,500 μg/ml chloramphenicol, 0-5% ethanol, or 0-1% chloroform.
The results showed that both strains had similar growth characteristics under most circumstances. Only when chloramphenicol was present in the medium, a difference in the ability to grow was observed. This difference was most pronounced at a concentration of 650 μg/ml chloramphenicol, where the absorbance of the wild-type culture after overnight growth was 1.7-fold higher than that of the PagL mutant (data not shown).
This suggests that PagL probably does not function in the adaptation to different growth temperatures or in the resistance against osmotic stress or organic solvents but that its activity affects the permeability of the outer membrane for hydrophobic compounds, such as chloramphenicol.

Discussion
A lipid A 3-O-deacylase, PagL, was originally identifi ed in S. Typhimurium (Trent et al., 2001a). Although similar activity was detected in some other bacteria, no homologs of pagL were identifi ed (Trent et al., 2001a). Now, we identifi ed pagL homologs in a range of Gram-negative bacteria, and we showed by cloning and expression of the corresponding genes in E. coli that at least two of them, those of P. aeruginosa and B.
bronchiseptica, are functional lipid A deacylases. The newly identifi ed PagL homologs are from organisms with genomes with a high GC content, and no other homologs were identifi ed in Enterobacteriaceae, which suggested that Salmonellae might have acquired the pagL gene by horizontal gene transfer from an organism with a high GC content. However, although all newly identifi ed pagL genes have a relatively high GC content (55-66%), pagL of S. Typhimurium has a GC content of only 39.5%, which is considerably lower than the S. Typhimurium average chromosomal GC content of 53% (McClelland et al., 2001). This observation suggests that S. Typhimurium has indeed acquired the pagL gene by horizontal gene transfer but not from an organism with a high GC content.
In previous work, it was demonstrated that outer membranes prepared from P. aeruginosa PAO1 harboured lipid A 3-O-deacylase activity (Basu et al., 1999). Furthermore, it is known that P. aeruginosa contains partially 3-O-deacylated lipid A species (Kulshin et al., 1991). The P. aeruginosa PagL homolog identifi ed here is most likely responsible for the 3-O-deacylase activity found previously in P. aeruginosa membranes. However, we did not fi nd an intact PagL homolog in the R. leguminosarum genome sequence, although 3-O-deacylase activity in its membranes has been described (Basu et al., 1999). BLAST searches with the unfi nished R. leguminosarum genome sequence did show the presence of two pagL homologs 2 ; however, the ORFs encoding the homologs were found to be disturbed by premature stop codons. Possibly, the available sequence still contains some errors. Also in other Gram-negative bacteria that are known to contain partially 3-O-deacylated lipid A, such as H. pylori (Bhat et al., 1994), no pagL homologs could be found. It is possible that the PagL homologs in these bacteria show only very low sequence similarity to the PagL family described in this paper, or 3-O-deacylation in these bacteria is mediated by entirely different  (Kol et al., 2003). Thus, the mere overproduction of PagL in our studies may already be suffi cient to induce phospholipids transport. This notion is consistent with the appearance of a slower migrating LPS form, even when inactive PagL proteins were expressed (Fig. 9B).
The high expression of the PagL homologs in E. coli allowed the determination of the processing sites by N-terminal sequence analysis. The identifi ed leader peptidase I cleavage sites for PagL (Pa) and PagL (Bb) correspond to the location predicted by the SignalP server (Nielsen et al., 1999). However, for PagL (St) , the cleavage site did corresponded neither to the SignalP predicted site (between amino acids 17 and 18 (AFA and CSA)) (Nielsen et al., 1999) nor to the position that was identifi ed in earlier work, in which the cleavage position was determined to be between residues 22 and 23 (AND and NVF) (Trent et al., 2001a). Because the -1 residue of the latter cleavage site does not conform to the consensus sequence S n XA (S n stands for an amino acid with a small neutral side chain), which is necessary for recognition by leader peptidase (von Heijne, 1983), this cleavage site is unlikely to be correct. The SignalP predicted site does match the consensus sequence for the -3 and -1 position, but the C domain, comprising the last six residues of the signal sequence does not include polar residues as one would expect it to do (von Heijne, 1983). The cleavage position identifi ed here between residues 20 and 21 (CSA and NDN) does match both criteria, with a cysteine at the -3 (S n ) position and a serine at -2, increasing the polarity of the C domain.
The poor conservation of the PagL sequence allowed us to speculate about the catalytic mechanism. Among the few totally conserved residues were a histidine and a serine, which are located in our topology model in a position similar to that of the active-site residues in OMPLA (Snijder et al., 1999), i.e., at the lipid-exposed side of a βstrand, close to the cell surface. By means of amino acid substitutions, we demonstrated that these conserved residues are indeed essential for catalytic activity. The position of the active-site residues at the lipid-exposed side of the β-barrel suggests that PagL, like OMPLA, may be active as a dimer to be able to form a substrate binding pocket.
Good candidates for the acidic component of the catalytic triad are the highly conserved aspartate and glutamate at positions 129 and 163 of the PagL (Pa) precursor sequence, respectively, which, in our model, are located in the β-strands fl anking the one containing the active-site His and Ser residues (Fig. 8). The Asp-129 is conserved in all homologs, except in S. Typhimurium, where it is replaced by glutamate. The Glu-163 varies more, with possible substitutions by an aspartate, or an asparagine. Although rare, asparagine can substitute for the acidic residue in serine hydrolases, as it does, for instance, in E.
coli OMPLA (Snijder et al., 1999). The physiological role of PagL-mediated deacylation of lipid A remains to be elucidated. The identifi cation of PagL homologs in a variety of Gram-negative bacteria, including non-pathogenic ones, such as P. putida and R. metallidurans, suggests that lipid A deacylation does not have a dedicated function in pathogenicity. To gain insight in its function, we tested whether pagL expression in P. aeruginosa can be infl uenced by the growth conditions. Supplementation of the growth medium with Mg 2+ ions did not affect pagL expression, consistent with the observation that 3-O-deacylase activity in P. aeruginosa membranes is not affected by the Mg 2+ concentration in the medium nor by a phoP mutation (Trent et al., 2001a). Also, other growth conditions tested did not appear to affect pagL expression. Furthermore, we compared the growth of wild-type P.
aeruginosa and a pagL mutant derivative under different conditions. The results showed a decreased resistance of the pagL mutant to chloramphenicol, suggesting that PagL activity, at least in P. aeruginosa, affects the permeability barrier of the outer membrane to hydrophobic compounds. The PagL family described here probably includes many more proteins because new BLAST searches just before submission of this paper (December 2004) using the newly identifi ed homologs as leads revealed several additional PagL homologs in the genomes of, for example, Agrobacterium tumefaciens (GenBank TM accession no. AAK87616), Methylobacillus fl agellatus (GenBank TM accession no. ZP_ 00173991), and Geobacter sulfurreducens (GenBank TM accession no. AAR36806).
Again, these proteins have low overall sequence similarity but a clearly conserved Cterminal region, including the histidine-serine couple that was identifi ed here as part of the active-site.